Structural Equation Modeling (SEM) 

Pre Requirements:- You will need data in a CSV as follows: 

# Install necessary packages (only once; comment out after installed)

install.packages("lavaan")          # Core SEM modeling package

install.packages("semPlot")         # For SEM diagram plotting

install.packages("lavaanPlot")      # Simple wrapper for creating SEM diagrams with lavaan

install.packages("DiagrammeRsvg")   # To convert graphs to SVG format

install.packages("rsvg")            # For exporting SVG to PNG

# Load libraries

library(lavaan)         # Core SEM modeling functions

library(semPlot)        # Plotting SEM models

library(lavaanPlot)     # Cleaner SEM plots

library(DiagrammeR)     # Graph rendering (used internally by lavaanPlot)

library(DiagrammeRsvg)  # Convert DiagrammeR object to SVG

library(rsvg)           # Convert SVG to PNG

# Load your SEM-ready dataset (should be a data frame with one row per site/observation)

sem_data <- read.csv("C:/Users/Rounak Choudhary/Desktop/env.csv")

# Check column names in your dataset

names(sem_data)

# Fit a full multiple linear regression model using all candidate predictors for Taxa

full_model <- lm(Taxa ~ AIV + Tree + Bare + Crop + Grass + IJI + Area, data = sem_data)

# Perform stepwise model selection to identify the best subset of predictors

best_model <- step(full_model, direction = "both")

# View the summary of the best selected linear model

summary(best_model)

# Define SEM model structure using lavaan syntax as a character string

model <- '

  Taxa ~ AIV + Tree + Bare + Area

'

# Fit the SEM model using lavaan

fit <- sem(model, data = sem_data)

# Show a detailed summary of the SEM model:

# - parameter estimates

# - fit statistics

# - standardized coefficients

summary(fit, fit.measures = TRUE, standardized = TRUE)

# Create a basic SEM plot from the model

lavaanPlot(

  model = fit,

  stand = TRUE,             # Show standardized coefficients

  coefs = TRUE,             # Display numeric coefficients

  covs = TRUE,              # Show covariances (if present)

  stars = "regress",        # Add significance stars to regression paths

  graph_options = list(rankdir = "LR")  # Left-to-right layout

)

# Create an enhanced SEM plot with custom node and edge styles

lavaanPlot(

  model = fit,

  stand = TRUE,

  coefs = TRUE,

  covs = TRUE,

  stars = "regress",

  node_options = list(         # Customize node appearance

    shape = "rectangle",

    color = "lightblue",

    fontcolor = "black",

    fontsize = 18,

    style = "filled",

    fixedsize = FALSE

  ),

  edge_options = list(         # Customize edge (arrow) appearance

    color = "gray20",

    fontcolor = "black",

    fontsize = 16,

    penwidth = 2

  ),

  graph_options = list(        # Customize graph layout and behavior

    rankdir = "LR",

    splines = "true",

    layout = "dot"

  )

)

# Store the SEM plot in a variable `p` as a DiagrammeR graph object

p <- lavaanPlot(

  model = fit,

  stand = TRUE,

  coefs = TRUE,

  covs = TRUE,

  stars = "regress",

  node_options = list(

    shape = "rectangle",

    color = "lightblue",

    fontcolor = "black",

    fontsize = 18,

    style = "filled",

    fixedsize = FALSE

  ),

  edge_options = list(

    color = "gray20",

    fontcolor = "black",

    fontsize = 16,

    penwidth = 2

  ),

  graph_options = list(

    rankdir = "LR",

    splines = "true",

    layout = "dot"

  )

)

# Convert the DiagrammeR graph to SVG format

svg_code <- DiagrammeRsvg::export_svg(p)

# Save the SVG as a high-resolution PNG image on your desktop

rsvg::rsvg_png(

  charToRaw(svg_code),

  "C:/Users/Rounak Choudhary/Desktop/sem_plot.png",

  width = 1600,

  height = 1000

)